Why are folks saying there is an AI bubble? A beginner’s quick briefing guide

Time to read:

4–5 minutes

You may have noticed that two things are happening at once.

People are using AI to do extraordinary things.

And people are investing amounts of money in AI that previously existed only in lawsuits involving royalty.

Naturally, you might wonder: is this a bubble?

To understand the concern, we need to talk about tokens, valuations and companies handing each other enormous cheques.

Stay with me. There will be a bakery.

1. Every time you ask AI something, somebody gets a bill

AI processes information in little chunks called tokens. These can be words, bits of words or individual characters. Both your question and the answer use them.

So when you ask ChatGPT to rewrite an email, that uses tokens.

When you say “make it friendlier”, that uses more.

When you say “not THAT friendly, he’s our accountant”, the meter keeps running.

Behind all this are computers using electricity and costing money.

Now imagine millions of people doing this, including someone asking an AI agent to build an entire business while they go downstairs for a biscuit.

The concern is simple: can AI companies charge enough to cover what people use, plus everything it costs to develop the technology?

A monthly subscription is a bit like an all-you-can-eat buffet. It works beautifully until a customer arrives with a wheelbarrow.

Making AI cheaper to run helps. But lower prices and heavier usage can eat into those savings. And serving customers is only part of the bill. Developing the next model and securing computing capacity also cost money.

2. The companies are priced as though the future has already happened

There are three different things here:

Revenue: money customers pay you.

Profit: what’s left after you pay the bills.

Valuation: what investors think your company is worth.

These are easily confused, particularly when someone would prefer you to focus on the largest number.

Imagine a bakery bringing in £100,000 a year.

Someone decides it’s worth £10 million because, in the future, everybody on Earth might need your particular muffin.

Those are made-up numbers. Here are some real ones.

In May 2026, Anthropic, the company behind Claude, raised funding at a valuation of $965 billion. Almost a trillion dollars. At the time, it reported an annual revenue “run rate” of $47 billion.

“Run rate” means what the company would bring in over a year if sales continued at that pace. It doesn’t mean it had already collected that much over twelve months.

Using those figures, investors valued Anthropic at roughly 21 times its annualised revenue.

Revenue, remember, is before paying for the computers, electricity, researchers and presumably quite a lot of coffee.

There is substantial growth behind the excitement. Anthropic reported $16.29 billion in revenue in the first half of 2026, compared with $4.59 billion for the whole of 2025.

So the customers are real. The money is real. The growth is real.

The worry is whether the company can eventually produce enough profit to justify that enormous price tag.

Selling muffins is encouraging.

Being priced as the future owner of breakfast is a bigger commitment.

3. Everybody appears to be everybody else’s customer

Here’s where you may want to get a pencil.

A big technology company invests in an AI company.

The AI company buys computing services from the big technology company.

The big technology company is also developing its own AI.

So it’s an investor, a supplier and a competitor. Presumably the Christmas party involves several lawyers and nobody leaving their laptop unattended.

Take Amazon and Anthropic.

Amazon announced another $5 billion investment in Anthropic, on top of $8 billion previously invested, with up to $20 billion more tied to future milestones.

Meanwhile, Anthropic committed to spending more than $100 billion on Amazon’s cloud technology over ten years.

These are different sums over different periods. It isn’t one cheque being passed straight back. But the relationship illustrates why people are looking closely at where the money goes.

Back to our bakery.

Your flour supplier invests in your business. You spend some of the investment buying their flour. They report flour sales. You report expansion.

Meanwhile, they also sell muffins.

None of this automatically means anything improper is happening. Real flour is being delivered. Real ovens are getting hot.

But eventually, enough people need to walk through the door and buy muffins to make the whole arrangement pay.

That’s the concern about circular financing. How much of the boom is supported by customer demand, and how much depends on these companies continuing to fund each other’s spending?

Does that mean AI is doomed?

No.

A useful technology can still attract wildly optimistic investment.

The internet changed the world. It also inspired investors to fund businesses whose entire strategy appeared to be “we have purchased a domain name”.

AI may become an enormous part of everyday life. That doesn’t guarantee every AI company will make money, or that every valuation makes sense.

The question is whether the profits will eventually catch up with the promises.

In the meantime, the electricity company would like to be paid in actual money.


Comments

One response to “Why are folks saying there is an AI bubble? A beginner’s quick briefing guide”

  1. My experience with these models shows a different pattern. I noticed a specific dataset size is the real factor here.

Leave a Reply

Your email address will not be published. Required fields are marked *